Method and System for Predicting Blood Flow Features Based on Medical Images

1. A computer-implemented method for automatically predicting a blood flow feature based on a medical image, comprising: acquiring, by a processor, image patches and a vessel related feature of a vessel tree; calculating, by the processor, the blood flow feature of the vessel tree using a learning network based on both the image patches and the vessel related feature of the vessel tree, wherein the learning network includes a multi-model neural network and a tree structure recurrent neural network connected in series.

2. The method of claim 1 , wherein the blood flow feature includes at least one of a fractional flow reserve, a blood flow, a blood flow rate, a micro-vascular resistance, and a blood flow pressure drop.

3. The method of claim 1 , wherein the vessel related feature includes at least one of an initial blood flow feature, a vessel structural feature, and a derivative feature, and each of the initial blood flow feature.

4. The method of claim 1 , wherein the vessel related feature of the vessel tree is obtained based on a 3D reconstruction of the medical image, by measurement, or by fluid simulation and calculation based on the medical image.

5. The method of claim 1 , wherein the multi-model neural network includes a convolutional neural network and a multi-layer neural network.

6. The method of claim 1 , wherein the image patches are 2D image patches or 3D image patches.

7. The method of claim 1 , wherein both the image patches and the vessel related feature are obtained along a centerline of the vessel tree.

8. The method of claim 1 , wherein, a tree-shaped structure of the tree structure recurrent neural network corresponds to a tree-shaped structure of the vessel tree.

9. The method of claim 1 , wherein, the tree structure recurrent neural network comprises multiple recurrent neural networks, each configured to calculate the blood flow feature of the vessel tree in a different direction.

10. A system for automatically predicting a blood flow feature based on a medical image, comprising: an acquisition interface configured to acquire medical images; and a processor configured to: reconstruct a 3D model of the vessel tree based on the medical images; acquire image patches and a vessel related feature of the vessel tree; calculate the blood flow feature of the vessel tree by using a learning network based on both the image patches and the vessel related feature of the vessel tree, wherein the learning network includes a multi-model neural network and a tree structure recurrent neural network connected in series.

11. The system of claim 10 , wherein the vessel related feature is obtained based on parameters of the 3D model of the vessel tree received from the reconstruction unit.

12. The system of claim 10 , wherein the processor is further configured to perform a fluid simulation calculation based on the reconstructed 3D model of the vessel tree to obtain an initial blood flow feature as the vessel related feature, the initial blood flow feature having a lower accuracy than that of the blood flow feature.

13. The system of claim 10 , wherein the processor is further configured to: acquire a training dataset including training image patches, training vessel related features of the vessel tree, and corresponding third blood flow features, the third blood flow features of the vessel tree being obtained by measurement; train the learning network by using the training dataset.

14. The system of claim 10 , wherein the multi-model neural network includes a convolutional neural network and a multi-layer neural network.

15. A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by a processor, perform a method for automatically predicting a blood flow feature based on a medical image, the method comprising: acquiring image patches and a vessel related feature of a vessel tree; calculating the blood flow feature of the vessel tree using a learning network based on both the image patches and the vessel related feature of the vessel tree, wherein the learning network includes a multi-model neural network and a tree structure recurrent neural network connected in series.

16. The non-transitory computer readable medium of claim 15 , wherein the blood flow feature includes at least one of a fractional flow reserve, a blood flow, a blood flow rate, a micro-vascular resistance, and a blood flow pressure drop.

17. The non-transitory computer readable medium of claim 15 , wherein the vessel related feature includes at least one of an initial blood flow feature, a vessel structural feature, and a derivative feature.

18. The non-transitory computer readable medium of claim 15 , wherein the vessel related feature of the vessel tree is obtained based on a 3D reconstruction of the medical image, by measurement, or by fluid simulation and calculation based on the medical image.

19. The non-transitory computer readable medium of claim 15 , wherein the multi-model neural network includes a convolutional neural network and a multi-layer neural network.

20. The non-transitory computer readable medium of claim 15 , wherein a tree-shaped structure of the tree structure recurrent neural network corresponds to a tree-shaped structure of the vessel tree.